test: cover CommunityPeople's unknown-section roster fallback end to end

[hivecommons-hive]

Test Improvement

src/components/CommunityPeople/index.js line 134 resolves its roster with
peopleData.people[section] || [], and nothing in a browser has ever taken the
right operand.

The published site renders the component from exactly two documents —
docs/community/technical-advisory-board.md:29 (section="tab") and
docs/community/index.md:133 (section="staff") — and
tests/community-people-data.test.mjs:92-104 fails the build when a section
rendered under docs/ is missing from data/community-people.json or carries
no entries. The fallback is therefore unreachable from published content by
construction. A data overlay cannot reach it either: emptying a section hands
the component an array straight off the object, while the || [] arm runs only
when the key is absent.

So the shape gets a page of its own, following
tests/e2e/fixtures/docs/cncf-project-card.mdx:

• tests/e2e/fixtures/docs/community-people.mdx (new) renders
  <CommunityPeople section="coverage-fixture-no-such-section" />.
  docusaurus.config.js:86-102 registers the e2e-coverage-fixtures docs
  instance only when E2E_COVERAGE=1, so npm run build:production, the
  gating "End-to-end tests" job and the deployed site never compile it.
  sectionsReferencedByDocs() walks docs/ only
  (tests/community-people-data.test.mjs:48), so the undefined section does
  not trip the contract above.
• tests/e2e/community-people-unknown-section.spec.js (new) asserts the
  grid from both ends: the published TAB page produces one grid with more than
  zero cards, and the fixture route produces one grid with zero. An empty grid
  on its own would also be produced by a selector that matched nothing, which
  is why the populated case comes first.

Evidence

• Unit — already covered. npm run test:unit:coverage (node v26.10.0,
  TZ=UTC) at main = 7ab301e: this file reports 100.00% lines / 100.00%
  regions, and tests/community-people.test.mjs:160 drives the arm directly.
• E2E — was not covered. CI artifact e2e-coverage, id 11383321728, run
  37395255849,
  job End-to-end coverage at 7ab301e:
  src/components/CommunityPeople/index.js | 100.00 | 87.10 | | 32 125 134.
• The arm is now taken in a browser. Verified locally at 7ab301e against
  npm run build:e2e:coverage:site: a coverage run containing the fixture case
  alone reports region 134 absent from this file's uncovered list.
• npm test (check + 1966 unit tests) passes on this branch.

What this does not do

Region 134 stays on the report's uncovered list in a run that also visits the
published page — the two instances' copies of the chunk map back to different
original coordinates, and getRegionCoverage in
tests/tools/e2e-coverage-report.mjs keys a region on its exact coordinates.
That is the attribution problem tracked in #1066 and #1079, the same limit
#1097 recorded for RadarReports; it is not something a fixture can reach.

Scope and coordination

This PR claims region 134 only of that row. Region 32 (the backdrop
onMouseDown arm) belongs to #1120 and region 125 (the onClose arrow) to the
#1079 class; neither is touched.

Two new files, zero edits. It does not touch
tests/e2e/fixtures/docs/cncf-project-card.mdx or
tests/e2e/cncf-project-card.spec.js (#1124),
tests/e2e/dialog-non-dismissing.spec.js (#1120),
tests/e2e-data-fixtures.test.mjs (#1110, #1113),
tests/tools/e2e-data-fixtures.cjs (#1110), or anything under
tests/e2e/fixtures/data/ or tests/e2e/fixtures/data-variants/. No workflow
file is involved.

Related Issue

Closes #1135

---

Human review required.

— hive: agent=quality backend=copilot model=claude-opus-5 copilot=1.0.88

[hivecommons-hive]

Important

Held for human review by the hive's ACMM level gate.

This PR was opened by the "quality" agent while Hive policy required a human checkpoint for that agent. Non-outreach agents are held at ACMM L3–L5; the outreach agent is always held because it publishes project-facing communication.

Hive will keep the hold label until a human removes it. Operators can make a deliberate one-off release during an ACMM level change with release_level_holds=true, but level changes never release this hold automatically.